Protein therapeutics are clinically developed and used as minorly engineered forms of their natural templates. This direct adoption of natural proteins in therapeutic contexts very frequently faces major challenges, including instability, poor solubility, and aggregation, which may result in undesired clinical outcomes. In contrast to classical protein engineering techniques, de novo protein design enables the introduction of radical sequence and structure manipulations, which can be used to address these challenges. In this work, we test the utility of two different design strategies to design novel granulopoietic proteins, using structural information from human granulocyte-colony stimulating factor (hG-CSF) as a template. The two strategies are: (1) An epitope rescaffolding where we migrate a tertiary structural epitope to simpler, idealised, proteins scaffolds (Fig. 1A-C), and (2) a topological refactoring strategy, where we change the protein fold by rearranging connections across the secondary structures and optimised the designed sequence of the new fold (Fig. 1A,D,E). Testing only eight designs, we obtained novel granulopoietic proteins that bind to the G-CSF receptor, have nanomolar activity in cell-based assays, and were highly thermostable and protease-resistant. NMR structure determination showed three designs to match their designed coordinates within less than 2.5 Å. While the designs possessed starkly different sequence and structure from the native G-CSF, they showed very specific activity in differentiating primary human haematopoietic stem cells into fully mature granulocytes. Morever, one design shows significant and specific activity in vivo in zebrafish and mice. These results are prospectively directing us to investigate the role of dimerisation geometry of G-GCSF receptor on activation magnitude and downstream signalling pathways. More broadly, the results also motivate our ongoing work on to design other heamatopoietic agents. In conclusion, our findings highlight the utility of computational protein design as a highly effective and guided means for discovering nover receptor modulators, and to obtain new mechanistic information about the target molecule.

Figure 1. Two different strategies to generate superfolding G-CSF designs. (A) X-ray structure of G-CSF (orange) bound to its cognate receptor (red) through its binding epitope (blue). According to the epitope rescaffolding strategy, (B) the critical binding epitope residues were disembodied and used as a geometric search query against the entire Protein Data Bank (PDB) to retrieve structurally compatible scaffolds. The top six compatible scaffolds structures are shown in cartoon representation. (C) The top two templates chosen for sequence design, were a de novo designed coiled-coil and a four-helix bundle with unknown function. The binding epitopes were grafted, and the scaffolds were optimised to rigidly host the guest epitope. (D-E) According to the topological refactoring strategy (D) the topology of the native G-CSF was rewired from around the fixed binding epitope, and then was further mutated to idealise the core residues (blue volume (E)) and residues distal from the binding epitope (orange crust (E)). Both strategies aimed at simplifying the topology, reducing the size, and rigidifying the bound epitope conformation through alternate means.


No relevant conflicts of interest to declare.

Author notes


Asterisk with author names denotes non-ASH members.